The H 2 SO 4 -HNO 3 -NH 3 system at high humidities and in fogs: 1. Spatial and temporal patterns in the San Joaquin Valley of California

,


Extremely high acidities have been reported in fogs and low stratus clouds collected in southern California
. The acidity of these fogs was due to H2SO, , and HNO 3, while NH 3 was found to be the main alkaline species of the order of 0.01-0.1 cm s -x I- Sehmel, 1980;Slinn, 1982].
The occurrence of fog enhances the removal of aerosol species because growth of particles to fog droplet size considerably .increases their deposition rates. Fog droplet deposition velocities of 2-7 cm s-x over short grass IDollard and Unsworth, 1983] and 2-4 cm s-• over a dirt surface [Waldman, 1986] have been reported.
The San Joaquin Valley of California is an excellent "field laboratory" for the interpretation of fog water composition in terms of the above processes. It is the site of major oil recovery operations, which release large amounts of SO2 and NO,•. In addition, agricultural and livestock-feeding activities provide important sources of NH3. Severe stagnation episodes, associated with persistent fog and low-lying stratus clouds, occur frequently during the winter months. These stagnation episodes are caused by persistent temperature inversions based a few hundred meters above the valley floor and below the surrounding mountain ridges [Holets and Swanson, 1981].
Tracer studies have documented the lack of ventilation in the valley during these prolonged episodes [Reible, 1982] In a preliminary study at a single site from December 1982 to January 1983, Jacob et al. [-1984a] documented the main A more comprehensive study was conducted during January 1984, of which this report presents the main results. Aerosol, fog water, and gas phase concentrations were monitored at a network of sites. Spatial and temporal variations of atmospheric concentrations are established that reflect both the geographical distribution of emission sources and the meteorological conditions. The concept of atmospheric alkalinity is introduced to predict the potential for high-acidity fog events. Pollutant accumulation and removal over the course of a stagnation episode are characterized under both foggy and nonfoggy conditions. The rate of oxidation of SO: to H•SO,• is estimated. The effect of stagnation on HNO 3 production is discussed. The partitioning of the H:SO,•-HNO3-NH 3 system between the gas phase, the aerosol, and the fog water is interpreted in a companion paper in terms of a thermodynamic model [Jacob et al., this issue]. The complete fog water, aerosol, and gas phase data set is given by Jacob [ 1985]. EXPERIMENTAL The volumetric concentrations of aerosol, HNO3(g), and NH3(g ) were monitored at eight sites over the period December 31, 1983, to January 14, 1984 (Figure 1). Samples were collected twice daily (0000-0400 and 1200-1600 PST) at the six valley sites and once daily (1000-1600 PST) at the two mountain sites (Tehachapi and Lake Isabella). The sampling stations were located on platforms 15 m above the ground (Wasco and Tehachapi), on the roof of a building or trailer (Bakersfield, Lost Hills, Buttonwillow, McKittrick, and Visalia), or on the ground, 1.

m above a grassy area (Lake Isabella). Two open-faced 47-mm Gelman Zefluor Teflon filters
(1-#m pore size) were operated side by side (10 L min -•) to provide duplicate determinations of the inorganic content of the aerosol. A flat cover 15 cm above the filters prevented collection of large particles by sedimentation. The Stokes number at the filter inlet is 0.05 for a 50-#m-diameter particle, so that even very large fog droplets should be efficiently sampled [Davies and Subari, 1982]. Under foggy conditions the determinations of total aerosol NO3-and NH,• + may be subject to errors because of evaporative losses (see the appendix), and those numbers subject to error were excluded from the data interpretation. A 47-mm Gelman Nylasorb nylon filter collected gaseous nitric acid immediately downstream of one of the Teflon filters, and an oxalic-acid-impregnated glass fiber filter collected gaseous ammonia immediately downstream of the other Teflon filter.
The filters were sealed in petri dishes and kept at 4øC following collection. The Teflon filters were extracted in 10 mL of distilled deionized water (Corning Megapure) for 90 min, using a reciprocating shaker; complete extraction was indicated by insignificant concentrations in repeated extractions. The extracts were analyzed for major ions using standard methods previously described by Munger et al. [1983]. The nylon filters were extracted for 90 min, using a reciprocating shaker in a solution 3 mM HCO3-and 2.4 mM CO3 :-(a conventional ion chromatography eluent). Oxalic-acid-impregnated filters were extracted and analyzed following the protocol of Russell [1983]. Ion chromatography revealed low levels of S(IV) in the Teflon filter extracts, but these concentrations were small compared to SO,• e-, and their contributions were ignored.
Although most of the aerosol sulfur is expected to be present as SO• •-, some of the measured SO,• •-may have resulted from the oxidation of reduced sulfur species on the filter or in the extract. Fogs were sampled by event at four sites ( Figure 1). Fog water samples were collected with a rotating arm collector [Jacob et al., 1984b] for intervals ranging from 30 min to 3 hours. The rotating arm collector has a theoretical sampling rate of 5 m 3 min-• and has been shown to collect fog water samples without evaporation or condensation. Laboratory calibration has indicated a lower size cut (50% collection efficiency) of 20-#m diameter. Because the instrument collects fog droplets by direct impaction and does not require drawing air through an inlet, large fog droplets are efficiently collected. Liquid water content in the fog was determined from the sampling rate of the instrument,• assuming that 60% of the total liquid water sampled was actually collected l- Waldman, 1986].
Fog water samples were preserved and analyzed for major ions and trace metals following the protocol described by Munger et al. 1-1983], with the exception described below. In fog water samples, significant S(IV) concentrations are found during conventional anion analysis by ion chromatography. A possible explanation is the formation of stable but reversible S(IV)-RCHO adducts, such as CH•(OH)SO 3- [Munger et al., 1984]. The standard ion chromatographic method (Dionex

AS-3 column, [3 mM HCO3--{-2.4 mM CO3 e-] eluent, 3 mL min -• flow rate) proposed by Dionex Corporation [1981]
and used by Munger et al. [1983] does not clearly separate S(IV) and NO 3 -. Better separation is achieved with a weaker eluent or with the Dionex AS-4 column, but quantification of S(IV) remains unsatisfactory. To solve this problem, aliquots for anion determination were spiked to 0.09 M H•Oe several minutes prior to injection and simultaneously made alkaline by the usual addition of HCO 3-and CO3 2-to match the eluent; this procedure was found to result in the quantitative oxidation of HSO 3 -and CH2(OH)SO 3 -standards to SO,• 2and the total suppression of the S(IV) peak in fog water samples. Sulfur(IV)concentrations were separately determined by a pararosaniline colorimetric method on aliquots preserved with buffered (pH 4) CH20 immediately upon sample collection [Dasgupta et al., 1980]. Fog water SO,• 2-concentrations were calculated by subtracting the S(IV) concentrations thus obtained from the SO,• 2-concentrations determined by ion chromatography. Concentrations of carboxylate ions were determined by ion exclusion chromatography [Keene et al., 1983].
To date there are no standardized sampling procedures for the collection of fog water and aerosol samples for chemical analysis. Therefore it is important to assess the errors associated with our methods. A detailed discussion of sampling biases, artifacts, and standard errors on our data is presented in the appendix.
Hourly To account for the equilibria between gas and aerosol phases, we define S(VI), N(V), and N (-III) to represent the element at the given oxidation state, both in the gas and aerosol phases. Thus N(V) includes HNO3(g) and NO3-, and N(--III) includes NH3(g) and NH,• +. We further define [A] as the concentration of constituent A in fog water (moles per liter of water), and (A) as the concentration of A in air (moles per cubic meter of air). Concentrations of HNO3(g ) will generally be given in equivalents, for consistency with the units of NO 3-and NH,• + concentrations. "Equivalent" in that sense refers to the proton donor/acceptor capacity of the gas when scavenged by the aerosol. Both HNO3(g ) and NH3(g ) contribute one equivalent per mole; 1 ppb = 43 neq m-3 at 5øC. In the first pattern (December 31 to January 1, January 10-12, and January 14), ground-based inversions formed by radiation at night and broke up the following afternoon, leading to mixing heights in excess of 1000 m above ground level (AGL). This pattern was usually associated with clear skies or high cloudiness, but fogs in ground-based inversions were occasional occurrences (for example, at Bakersfield on December 31). Figure 3a gives the average wind vectors on the days when this pattern was observed. A net slow NW flow was observed on the valley floor, and upper level winds were NW; this is the usual flow in the area and reflects the circulation around the Pacific High off the California coast. Terrain influences in the southern end of the valley led to convergence of the flow in the SE corner of the valley. Surface winds in the valley frequently shifted in direction, and erratic low winds were typically observed under nighttime stable conditions. Concentrations of trace gases at Kernridge were lowest on the days when this first pattern was observed ( Figure 2) and so were aerosol concentrations [Jacob, 1985]. Because net horizontal transport was very slow (Figure 3a), it is unlikely that surface winds ventilated the valley by transport over the mountain ridges in the SE corner of the valley. Aerosol concentrations at Tehachapi and Lake Isabella remained much lower than in the valley, which is evidence against such transport. Instead, pollutant removal was due to rapid vertical mixing as the inversion broke up in the afternoon; this vertical mixing diluted the polluted air parcels and allowed their rapid transport by strong upper level NW winds to the surrounding air basins.

WEATHER PATTERN AND POLLUTANT TRANSPORT
A different mixing height pattern was observed on January 2-7; during that period a strong temperature inversion based a few hundred meters above the ground persisted over the valley. This inversion was due to mesoscale subsidence associated with a stationary high-pressure system (Great Basin High) centered over southern Idaho and northern Nevada. Upper level winds at Edwards Air Force Base switched from NW to east during that period (circulation around the Great Basin High), and afternoon winds at Tehachapi were SE. The valley was capped throughout the January 2-7 period by a stratus cloud filling the upper part of the mixed layer (

AVERAGE $(VI), N(V), AND N(--III) CONCENTRATIONS AT EACH SITE
The SSJV is the site of important SO2, NOx, and NH 3 emissions (Table 1). The geographical distribution of the main emission sources is shown in Figure 1. Most of the NH 3 is emitted from confined-feeding operations concentrated on the east side of the valley, especially around Bakersfield and Visalia. Another important source of NH 3 is cropland, which occupies most of the land in the valley floor not used for oil recovery operations, and the associated fertilizer use. Emissions of SO2 and NO,, are concentrated in the east side and west side oil fields of the SSJV; the oil field emissions originate mostly from small boilers, which release their exhausts 10-20 m above the ground and therefore affect the immediate surroundings. Mobile sources (two major highways, off-road farm equipment, and city traffic) also contribute to NOx emissions. Spatial patterns of aerosol and fog water concentrations (Tables 2 and 3a-3c) directly reflected the distribution of emission sources. The ionic content of aerosol in the valley was dominated by SO½ 2-, NO3-, and NH½ +, which typically contributed over 90% of the total measured ionic loading.

Concentrations of N(-III) were highest at Bakersfield and Visalia, near the large cattle feedlots, and lowest on the west side (Lost Hills and McKittrick). Mountain sites (Lake Isabella and Tehachapi), which have no important local NH 3 sources, had very low N(-III) concentrations.
Concentrations of S(VI) were highest at Bakersfield, which is within the east side oil fields; they were also high at McKittrick, which is within the west side oil fields, and at Buttonwillow, directly downwind of the west side oil fields. Concentrations of S(VI) at Wasco, Lost Hills, and Visalia were lower, reflecting their respective distances from oil recovery operations. Concentrations of S(VI) at the two mountain sites were very low and indicated no observable impact from the valley air.
Because most of the NO,, was emitted from the same sources as SO2, one would expect the spatial distribution of N(V) concentrations to be similar to that of S(VI). Indeed, N(V) concentrations were highest at Bakersfield; however, N(V) concentrations at McKittrick were low. The lack of NH 3 at McKittrick frequently led to acidic conditions, in which N(V) would be mostly present as HNO3(g) and therefore quickly removed by deposition. This point will be addressed in more detail below. Concentrations of N(V) were higher at Visalia than would be expected from the spatial distribution of S(VI); Visalia is an important population center, and emissions from mobile sources were probably the dominant source of N(V) precursors at that site.
Fog water concentrations of trace metals were consistent with the above analysis (Table 3b). Concentrations of Ni and V, which are almost exclusively associated with residual oil burning [Cooper and Watson, 1980], were high at Bakersfield,   (Table 2) the reference pH is of the order of 4.5. This is still above the threshold at which environmental damage from "acid fog" may be anticipated [Scherbatskoy and Klein, 1983;Granett and Musselman, 1984;Hoffmann, 1984].

McKittrick, and Buttonwillow and low at Visalia. Concentrations of Pb, which is emitted by automobile exhaust
Fog water with [ALK] < 0 is said to contain inorganic acidity, and its pH is lower than that of the reference system. Such fog water has zero buffer capacity with respect to further inputs of strong acids, which may then lead to extremely acidic conditions. We will use the presence of inorganic acidity as an operational definition of the term "acid fog." Alkaline fog water (defined by [ALK] > 0) has a pH higher than that of the reference system and will neutralize acid inputs until exhaustion of the alkalinity. However, [ALK] is not a true measure of the acid-neutralizing capacity of the fog water because it ignores exchanges with the gas phase. In particular, alkaline fog water may support a substantial NH 3 vapor pressure [Jacob et al., this issue], which provides an additional source of acid-neutralizing capacity to the fog water. Further, fog water alkalinity depends on the fraction of the aerosol scavenged in the fog. To use alkalinity as a measure of the acid-neutralizing capacity of the atmosphere with respect to acid fog, we need to introduce a more general concept, atmospheric alkalinity.
The atmospheric alkalinity (ALK) (equivalents per cubic meter of air) is the sum of total aerosol alkalinity and gas phase alkalinity in an atmosphere. That atmosphere may or may not contain fog. The reference system is the same as that used for defining fog water alkalinity; gas phase CO: and other weak acids are reference species. The aerosol alkalinity (ALK)a is given by (1), where fog water concentrations (moles per liter of water) are replaced by total aerosol concentrations (moles per cubic meter of air). The gas phase alkalinity is given by the proton donor and acceptor capacities of watersoluble atmospheric gases with respect to the reference system. Atmospheric alkalinity is a conserved quantity upon fog formation and can therefore be used to predict the potential for acid fog from aerosol and gas phase measurements taken under nonfoggy conditions. Atmospheric alkalinity also gives the amount of strong acids that may be emitted to the atmo- Since we have argued that the inorganic acidity in fog is mostly controlled by species in the HeSO,•-HNO3-NH 3 system, we will assume that the only gas phase contributors to (ALK) are HNO3(g) and NH3(g). A calculation of the ultimate alkalinity of an air parcel should include SO2(g); however, the contribution of SO2(g) to (ALK) may be limited by the slow rate of SO2 scavenging by fog . For now we ignored the SO2(g) contribution and calculated the atmospheric alkalinity from the expression: (ALK) -(ALK)a + (NH3(g)) -(HNO3(g)) Average alkalinities at each site are given in Table 4a. Because of the possibility of HNO 3 or NH 3 volatilization from aerosol filter samples collected in fog (see appendix), (ALK) and (ALK), were calculated only for nonfoggy conditions. (ALK) and (ALK), were usually small numbers determined by the difference of two large numbers, so the standard errors were fairly large. The calculation of the fog water alkalinity [ALK-I involved subtracting a small number from a large number; the resulting standard errors were small and were not indicated explicitly.  The western edge of the SSJV currently suffers from a general acid fog problem, as shown by the negative average values of (ALK). In the remainder of the valley, (ALK) > 0, and fog water is not usually acidic. However, the average values of (ALK) at Bakersfield and Wasco presently amount to less than 20% of (S(VI) + N(V)) equivalent concentrations. If NH3 emissions in the SSJV decrease by 20% compared to their current level, for example, because of a decline of the cattle industry or fluctuations in the soil moisture and temperature, a general acid fog situation in the east side of the SSJV will result. The same result will be achieved by a 20% increase in

(S(VI) + N(V)) equivalent concentrations due to a rise in SO2 and NO,, emissions. Fog water alkalinity at Visalia will not be affected by these changes in SSJV emissions, considering the large (ALK)/(S(VI) + N(V)) ratio at that site; therefore there is little risk that an acid fog problem in the SSJV could spread to the northern part of the San Joaquin Valley. The partitioning of the atmospheric alkalinity between the gas phase and the aerosol is of interest. Scavenging of NH3(g)
to form ammonium salts of weak acids could be a source of important alkalinity in the aerosol or fog water. Indeed, significant alkalinities were found in fog water; however, aerosol collected under nonfoggy conditions was never significantly alkaline. Although the error bars on the determinations of (ALK)a were large, the absence of positive (ALK)a values in the presence of large excesses of NH3(g), as at Visalia, strongly suggests that the alkaline ammonium salts are volatile under nonfoggy conditions. This hypothesis is supported by concurrent sampling of aerosol and fog water at Visalia, where  aAverage liquid water content for the subset of fog water samples.

NH½ + was in excess of NO 3 -and SO4 2-in the fog water but not in the dried aerosol. Artifact aerosol neutralization should not occur during filter storage (see appendix). Under nonfoggy acidic conditions the aerosol contained significant inorganic acidity when S(VI) was present in excess of N(-III). This occurred in six of the samples, all at McKittrick. In the remainder of the samples collected under acidic conditions the aerosol was neutralized, and the inorganic acid-
bFrom January 5 to January 14.
'From December 31 to January 7. Formic and acetic acids are efficiently scavenged in fog water at pH > 5' they are highly soluble, as indicated by their large Henry's law constants (HHcoOH,298 = 3.7 x 103 M atm-•, HCH3COOH,298 = 8.8 x 10 3 M atm-• [Weast, 1984]), and they are mostly dissociated, as indicated by their acidity constants (see above). At Visalia the fog water pH was much higher than at Bakersfield or Buttonwillow, and the contribution from HCO 3-to [ALK] was correspondingly larger. Carboxylate anions did not provide higher contributions at Visalia than at Bakersfield or Buttonwillow; since carboxylic acids are already efficiently scavenged at pH 5-6, raising the pH higher leads to little additional scavenging.

The goal of this section is to interpret the accumulation of S(VI), N(V), and N(-III)
species over the course of the January 2-7 severe stagnation episode, in terms of atmospheric production and removal mechanisms. As shown in Figure 5, the stagnation episode was generally associated with high concentrations of S(VI), N(V), and N(-III).
The inversion base was roughly stable at h = 400 m AGL throughout the episode, and the residence time for air parcels in the SSJV was '•a--5 days. Therefore deposition was a more important removal pathway than ventilation for species with deposition velocities  centrations, probably due to rapid deposition of fog droplets. Jacob et al. [1984a] have previously suggested that enhanced aerosol deposition in fogs efficiently limits pollutant accumulation during stagnation episodes, and our data support this hypothesis. Further decreases in atmospheric concentrations were observed on January 7-8, due to the rise of the inversion base ( Figure 2) and deposition from drizzle. As S(VI) and N(V) were produced over the course of the stagnation episode, significant inorganic acidities were observed at Wasco, Lost Hills, and McKittrick. At Bakersfield, sufficient N (-III) was available to totally neutralize acid inputs. Because of the remarkably stable mixing height and the lack of ventilation we can attempt to apply stirred-tank considerations to calculate the rates of H2SO½ and HNO 3 production in the SSJV. The residence time z of a species in a stirred tank is given by the expression: where v is the deposition velocity and k is a first-order chemical loss rate. Of special interest is the period January 2-5, ranging from the onset of stagnation to the first widespread valley fog. The SSJV floor (Bakersfield, Wasco, and Lost Hills) remained overcast throughout that period, and roughly constant values of v and k may be expected. We can thus follow the evolution over 4 days of a very well-controlled stagnant atmospheric system.

Production of H2S0½
Concentrations of S(VI) increased progressively on the SSJV floor during the nonfoggy January 2-5 period. Concentrations remained very low at Visalia, which is out of the SSJV and far from SO2 sources. Sulfur dioxide, the main precursor of S(VI) in the SSJV, has a deposition velocity of the order of 1 cm s-• over grass I -Sehmel, 1980]. A steady state for SO2 in the SSJV should therefore be approached on a time scale of the order of 1 day after the onset of stagnation, and this appeared to be the case at Kernridge (Figure 2). Measured concentrations of SO2 at Bakersfield, Kernridge, and Lost Hills during January 3-5 averaged 25, 12, and 3 ppb, respectively' a strong spatial gradient in SO2 concentrations was maintained because mixing was slow. Modeling of SO2 transport under stagnant conditions [Aerovironment, Incorporated, 1984] indicates that the SO2 concentration field in the SSJV should be bounded on the lower end at Wasco and Lost Hills and on the upper end at Bakersfield. Therefore we expect steady state SO2 concentrations in the SSJV mixed layer to range between 3 and 25 ppb. A stirred-tank calculation based on the volume of the SSJV mixed layer and the emission data of Table 1, assuming chemical loss to be slower than deposition (Vso2 = 1 cm s-a), gives an average steady state SO2 concentration of the order of 9 ppb in the SSJV. This is consistent with our observations. The accumulation pattern of S(VI) on the SSJV floor was consistent with a pseudo first-order conversion rate of SO2 to S(VI). Concentrations of S(VI) increased relatively steadily during the January 3-5 period, when SO2 concentrations were approaching steady state. This steady conversion of SO2 maintained important differences in S(VI) concentrations from site to site within the SSJV. A time lag for S(VI) production was clearly seen at Bakersfield on January 2, attributable to the time required for SO2 to accumulate after the onset of stagnation. The profiles of S(VI) concentrations in the SSJV did not suggest an approach of steady state by January 5; the residence time of S(VI) aerosol in the SSJV under nonfoggy conditions was thus longer than 3 days, indicating a deposition velocity Vs•w• < 0.05 cm s -• (equation (5)). This is in agreement with predicted deposition velocities for particles in the 0.05-to 1-#m size range at low wind velocities [Sehrnel, 1980]. Over the period January 3-5, S(VI) was produced in the SSJV much faster than it was removed, and we can to a first approximation equate the observed rate of S(VI) accumulation to the rate of SO2 conversion. The average rates of
Because of the lack of photochemical activity during the stagnation episode (see Figure 6 and discussion below), conversion of SO2 to H2SO ½ must have proceeded predominantly in the aerosol and the cloud droplets. A likely pathway is metal-catalyzed autoxidation in the aqueous phase , which does not require photochemically generated oxidants. Considering that a stratus cloud filled a large fraction of the mixed layer during the period of January 3-5, an important question is to determine if the principal site for S(IV) oxidation was the cloud or the haze aerosol below. Concentrations of SO2 progressively decreased at Kernridge during the foggy January 5-7 period, while concentrations of CO and NOx (which was emitted from the same sources as SO2) kept on increasing. This is strong evidence that removal of SO2 from the atmosphere was enhanced in the presence of fog. From Figure 2 the rate of SO2 scavenging by fog appears to be on the order of 5% h-1. This enhanced scavenging of SO2 in fog does not necessarily imply enhanced production of S(VI); S(IV) may be stabilized in the aqueous phase by formation of adducts [Munger et al., 1983[Munger et al., , 1984 or removed by deposition before being oxidized. We tried to evaluate S(VI) production directly in fog droplets by comparing S(VI) concentrations in successive fog water samples collected at one site, using Ni and V as conservative tracers for sulfur. However, we did not obtain statistically significant rates of SO2 conversion in fog water (2 q-6% h-• at Bakersfield, 0 q-3% h -a at McKittrick [see Jacob, 1985]). This failure to find statistically significant rates is probably due to the complex total N(V) concentrations. This indicates displacement of NO 3 -by H2SO4 in the aerosol, followed by rapid deposition of HNO3(g). The atmospheric lifetime for HNO3(g ) is thus short (<0.5 days), which implies a large deposition velocity (> 1 cm s-•).

Accumulation of NH 3
Almost all of the NH 3 emitted in the SSJV during the stagnation episode was used to neutralize acid inputs. The fate of the resulting (SO½ 2, NO3-, NH½ +) aerosol has been discussed above. At Visalia, however, acid inputs were small, and a large fraction of total N(-III) remained in the gas phase as NH3(g).
No accumulation of NH3(g) was apparent at that site over the course of the episode, and this suggests an atmospheric residence time of <0.5 days for NH3(g ) (deposition velocity of > 1 cm s-•). Frequent fog and drizzle after January 5 at Visalia resulted in an important depletion of NH½ + aerosol, but NH3(g ) concentrations were unaffected. Ammonia is poorly scavenged at the high pH values typical of Visalia fog water [Jacob et al., this issue].

Production of HN03
In addition to H2SO,•, HNO3 was produced. However, N(V) did not accumulate as steadily as S(VI). Concentrations of N(V) in the SSJV increased rapidly at the onset of stagnation (January 2-3) but did not increase after January 3. Nitrogen oxides accumulated steadily throughout the stagnation episode (Figure 2), showing no indication of loss from chemical conversion. Concentrations of N(V) did not show the large differences from site to site that were observed for S(VI), even though SO2 and NO,, mostly originated from the same combustion sources.
A logical explanation for these observations is that the rate of HNO 3 production was slow during the stagnation episode because of the widespread and persistent low overcast. Figure  6 shows the NO and 03 concentration profiles at Kernridge. Ozone levels prior to the onset of stagnation were relatively high, indicating substantial photochemical activity; HNO 3 production by the reaction NO: + OH should proceed rapidly under those conditions. Also, 03 was always present in excess of NO, so that HNO 3 could be produced at night by heterogeneous pathways initiated by the reaction NOe + 03 [-Heikes and Thompson, 1983]. After the onset of stagnation, however, the overcast restricted photochemical activity, and very low 03 concentrations were observed; OH concentrations were probably very low. Further, since NO progressively accumulated to levels sufficient to titrate 03 fully, the reaction NO2 + 03 did not proceed. Therefore little secondary production of HNO3 would be expected in the SSJV after January 3. Fog did not perceptibly enhance the conversion of NO,, to HNO3; concentrations of NO,, at Kernridge kept on increasing during the January 5-7 foggy period, similarly to CO, which is not water-soluble. This is consistent with the poor solubilities of NO and NO2 in water at atmospheric concentrations [Schwartz and White, 1981].
The NO3-aerosol present after January 3 was therefore mostly aged aerosol, slowly mixing within the SSJV. As mixing proceeded, the differeni:es in NO3-concentrations from site to site became progressively weaker. Concentrations of N(V) at Bakersfield and Wasco were similar on January 4-5, even though S(VI) concentrations were much higher at Bakersfield. At Lost Hills and McKittrick, acidification of the atmosphere coincided with a brief increase in HNO3(g) concentrations, immediately followed by an important drop in

Stirred-Tank Simulation
The above discussions have shown that the profiles of concentrations versus time during a stagnation episode can be successfully interpreted, based on stirred-tank considerations of pollutant accumulation and removal; however, the differences in concentrations from site to site clearly indicate that a stirred-tank model for the SSJV as a whole is not an adequate modeling tool. The major reason is that internal mixing is slow. Nevertheless, the success of the stirred-tank model in interpreting the data at individual sites suggests that one could model the SSJV by subdividing it into a number of cells where the stirred-tank approximation could be properly invoked. Such an exercise is beyond the scope of this paper; however, for the sake of illustrating and summarizing our discussion of field data we will present the results of a stirredtank calculation applied to the entire SSJV. The accumulation of constituent A in such a model is described by

-• [(Aa)va, + (Aa)vaa + (Af)vA, ] -(A)/z a (6)
where Aa, Aa, and Ay are the gas phase, aerosol phase, and fog water phase species, respectively, E z is the emission rate averaged over the volume of the mixed layer (moles per cubic meter per day), and k' is the pseudo first-order rate for conversion of precursor B to A. We simultaneously solved the coupled stirred-tank equations for SO 2, NOx, S(VI), N(V), and N(-III).
The model conditions are given in Table 5 and are for the most part deduced from our discussion of the field data. The emission dater are those of Table 1, averaged over the volume of the SSJV mixed layer. We assumed that the aerosol was a neutralized mixture under nonfoggy conditions if N(-III) was in excess of S(VI), and that the formation of NH4NO 3 aerosol was sufficiently favored to prevent HNO3 and NH3 from coexisting in the gas phase under any condition (valid if the atmosphere is sufficiently humid). The simulation was run for 4 days from the beginning of the episode, with a foggy period extending from t = 2 days to t = 3 days. On the basis of results presented by Jacob [1985], Jacob et al.  Figure 7 shows the predicted concentration profiles. Aerosol accumulates rapidly under nonfoggy conditions and is partially removed by fog. The main features of the observed concentration profiles are reproduced, in particular, the accumulation patterns for SO4 2-and NO3-. The emission rates of Table 1 lead to concentrations of species that are in the range of those observed. Some excess alkalinity as NH3(g) remains present throughout the episode. CONCLUSION A systematic characterization of the H2SO4-HNO3-NH3 system in the fog water, the aerosol, and the gas phase was conducted at a network of sites in the San Joaquin Valley of California. Spatial patterns of atmospheric concentrations reflected the geographic distribution of oil recovery operations (SO2, NO,,) and livestock-feeding and agricultural activities (NH3). The acidity of the fog water was found to be determined by the relative abundances of local acidic (SO2, NO,•) and alkaline (NH3) emissions. A region of prevailing acidic conditions was identified on the western edge of the valley, where NH3 emissions were low. Elsewhere, sufficient NH3 was available to fully titrate the acidity. In the southern end of the valley, where major oil recovery operations release large amounts of SO2 and NO,o a precarious atmospheric balance was found between high concentrations of acids and bases.
The concept of atmospheric alkalinity was introduced as a quantitative measure of the acid-neutralizing capacity of the atmosphere with respect to fog. On the basis of this concept we predicted the regional potentials for high-acidity fog events in the San Joaquin Valley. We concluded that small changes in the activities of the agricultural industry or the oil industry could lead to widespread "acid fog" in the southern end of the valley.   > 1 cm s-x). Decreases in aerosol concentrations were observed following fogs and were attributed to the rapid deposition of fog droplets. Therefore the occurrence of fog was found to effectively limit pollutant accumulation during stagnation episodes. Secondary production of strong acids under stagnant conditions entirely titrated available alkalinities at the sites farthest from NH3 emissions. A steady conversion rate of SO2 to H•.SO½ was estimated at 0.4-1.1% h -x under overcast stagnant conditions. Conversion of NO,, to HNO 3 was rapid at the beginning of the episode but dropped as the widespread and persistent low overcast reduced photochemical activity. Removal of SO2 was found to be enhanced in fog, compared to nonfoggy conditions, but NO,, was not scavenged in fog. Acidification of the atmosphere was associated with a brief increase in HNO3(g) followed by a drop in total N(V) concentrations; this was explained by the displacement of aerosol NO3-by H2SO,•, followed by rapid deposition of HNO3(g).

Fog Water Concentrations
A recent intercomparison study of fog water collectors [Hering and Blumenthal, 1985;Waldman, 1986] has demonstrated that the California Institute of Technology rotating arm collector provides reproducible and representative samples under both light and heavy fog conditions. The uncertainty on the concentrations of major ions (determined from samples collected with two rotating arm collectors set side by side) was found to be about 15%. Errors due to chemical analysis in the laboratory were about 5% for all analyzed ions. No significant differences in ionic concentrations were found between samples collected concurrently with the rotating arm collector, a jet impactor [Katz, 1980], and a screen collector [Brewer et al., 1983], set side by side.
Ionic balances are an indicator of whether all ionic components in the sample have been accounted for in analysis. Ionic balances in the fog water samples were 0.98 _+ 0.18 at Bakersfield (n--15), 1.02 + 0.09 at McKittrick (n = 53), 1.11 __+ 0.18 at Buttonwillow (n--7), and 0.60 + 0.17 at Visalia (n = 12). The ionic balances were calculated from the following molar ratio' where S(IV) was assumed to be monovalent ]. We have argued previously (equation (2)) that this is a good assumption even at high pH. The most reliable ionic balances were found at McKittrick, while at Visalia there was a considerable and consistent anion deficiency. Fog water at Visalia had a consistently high pH and as a consequence could contain important alkalinity; therefore additional weak acid anions must be considered in an ionic balance. In fog water with low pH, such as at McKittrick, weak acids are mostly present in undissociated form and thus (A1) represents a good balance of cations to anions.
The principal ionic contributors to alkalinity in the pH range 3-8 are expected to be HCO 3-and carboxylate ions. Fog water concentrations of four carboxylate ions were determined on a subset of the fog water data set (Table 3

Concentrations
The filter methods used in this study may lead to two major types of error: (1) N(V) and N(-III) artifacts, and (2) random errors from the sampling process. We will address each in order. An additional source of error could be the neutralization of acidic and alkaline aerosol by absorption of NH3(g ) and HNO3(g), respectively, during filter storage; however, blank oxalic-acid-impregnated filters and nylon filters were found to remain blank even after extended storage. Therefore aerosol neutralization during storage did not seem to occur. Stelson and Seinfeld [1982] have shown that an increase in temperature at constant dew point during sampling may voltatilize ammonium nitrate collected on Teflon filters and result in artifact HNO3(g) and NH3(g ). Further, absorption of gaseous nitric acid on the Teflon filter may result in artifact aerosol nitrate [Spicer and Schumacher, 1979;Appel et al., 1980]. In an intercomparison study of gaseous nitric acid measurement methods, Spicer et al. [1982] found good agreement between the dual-filter method (used here) and other methods. Further, they found that the dual-filter method was accurate in measuring total N(V).
The study of Spicer et al. [1982] was conducted under hot, dry conditions. Potential biases are different under the cool, humid conditions found in the San Joaquin Valley. Nighttime samples (0000-0400 PST) were probably unaffected by volatil-ization because temperatures during the sampling period either remained constant or decreased. During the day, temperature changes were usually small because of the overcast conditions' temperatures recorded hourly at Bakersfield between 1200 and 1600 PST increased on only five of the 15 sampling days and never increased by more than iøC, except on January 14 (when a 3øC increase was observed). Still, an increase of 1øC in temperature at constant dew point may increase the dissociation constant K-P-•o3 x P•,3 by a factor of 2 under high-humidity conditions [Stelson and $einfeld, 1982]. The formation of aerosol ammonium nitrate is strongly favored thermodynamically, and aerosol concentrations could not be significantly affected by volatilization; on the other hand, high relative errors may occur in the determination of the gas present at the lowest concentration.
Since that concentration was often near or below the detection limit, the error was of little consequence.
Teflon filters run in dense fogs accumulated drops of liquid water at the surface. In those particular cases the filters were dried in the open before being sealed. Nitric acid scavenged in acidic fog volatilizes during drying, leading to an underestimate of total aerosol NO 3 -. No significant NO 3 -loss should occur in nonacidic fog because in that case NO 3-remains in the aerosol phase as the fog dissipates. Similarly, NH4 + aerosol should not volatilize in acidic fog. Volatilization of NH,• + from filters collected in alkaline fog depends on the stability of the ammonium salts of weak acids, which appear to be volatile; some volatilization of NH,• + was found to occur in a sample collected during fog at Visalia [Jacob, 1985].
"Random errors" are of three types' (1) uncertainty in the flow rate through the filter, (2) uncertainty in the efficiency of recovery by extraction, and (3) analytical error. Because the first two sources of error affect the sample as a whole, we expect a correlation to exist between the errors on the different species. To test for these errors, concentrations of SO,, 2-, NO3-, and NH,• + were determined in duplicate for n-45 pairs of filter samples "1" and "2," collected side by side. Concentrations of C1-were also determined, but the errors on those concentrations were generally lower than the C1-filter No duplicate analyses were made for Na +, K +, Ca 2+, and Mg 2 +. Replicate analyses of standards indicate a,•,x 2 of about 5% for these four ions. Duplicates for NH3(g ) and HNO3(g) determinations were not collected, but the trA.x 2 values should be the same as for NH,• + and NO3-, respectively. In addition to the analytical error, concentrations of all constituents were assumed to be subject to the same error a•,. This assumption implies that the variability of the flow rate through the filter is the major contributor to a•,, which seems justified, since filter extraction efficiencies are better than 95% [Russell and Cass, 1984].
The analytical detection limits for a 4-hour sample correspond to 4 neq m-3 for NO 3 -, SO,• 2-, HNO3(g), and cations other than NH,• +, and 8 neq m -3 for NH,• +. Filter blanks for all constituents except CI-and NHa(g) were below these detection limits. Because of substantial filter blanks, effective detection limits for NH3(g ) and C1-for a 4-hour sample were 17 neq m-3 and 20 neq m-3, respectively.